Outline of the talk

• The Benefits of Automation
• The Real Benefits of Automation
• How FAVE works
• The FUD Surrounding Automation
• eVoweluate

Some Undeniable Benefits

• It’s a bit faster.
• You get more, and richer data.

Dinkin (2009)

• 57,464 formant measurements
• 119 speakers
• Average of 483 mesurements per speaker

Philadelphia Neighborhood Corpus

• 743,802 formant measurements
• 397 speakers
• Average of 1,874 measurements per speaker
• Average of 0.94 vowels per second.

More Data

/ay/ followed by different /t,d/ contexts.

More Data

More Data

Richer Data

Richer Data

Richer Data

Reproducibility

Researcher Heuristics

• Choosing a measurement point.
• Adjusting the LPC settings.
• Choosing vowels to measure or ignore.

Researcher Effects

• How explicitly defined their heuristics are.
• How strictly they enforce them.
• How experienced and skilled they are.
• Their recent caffiene consumption, and quality of previous night’s sleep.

Automation

Knowable, Explicitly Defined, Exceptionless

• Measurement point selection method.
• LPC parameter setting method.
• Decision process for measuring a vowel or not.

Automation

Eliminated

• Researcher experience and skill.
• …oops

Reproducibility

Reproducibility

FAVE

FAVE

• For each vowel token, the F1 and F2 estimates you could get for different LPC parameter settings constitute a candidate set.
• Choose a winner based on its multivariate distance (based on F1, F2, log(B1), log(B2)) to the Atlas of North American English’s distribution for that vowel class.
• Logic: If there is an LPC setting whch is produces a measurement close to the ANAE distribution for that vowel class, it’s probably ok.

FAVE - Once more, but Bayesian this time

• The ANAE distribution for a vowel class is the prior.
• The candidate set of potential formant estimates is the likelihood.
• The winner is the posterior.

• Like most worries about Bayesian reasoning, people worry that the prior might exert too strong an influence on the posterior.
• Fortunately, the prior’s influence here doesn’t seem to be too strong.

FAVE - Step 1

FAVE - Remeasurement

FAVE - Step 2

FAVE - Results

FAVE - Future Directions

• feature/iterremeasure
  • Continue iterating through re-estimation either until it arrives at a stable distribution, or a maximum iteration is reached.
• feature/bootstrap
  • Try to figure out the best LPC setting based on the distribution of the candidate set alone.
  • This approach would only be appropriate for vowel classes with a lot of tokens.

FAVE - Future Directions

Errors

People make errors too

The “correct” answer isn’t well defined anyway

“Black Box”

People are black boxes

FAVE is not a black box

It’s like…

It’ll encourage…

Moral Panic

• Either we’re clever enough to learn how to appropriately use new research tools, or we’re not.
• Automation is not an excuse to understand your data less well.

• The proof is in the pudding.

eVoweluate Survey